System identification based on Hammerstein model

From MaRDI portal





The Hammerstein system identification is considered. A scheme yielding the estimates of the linear dynamics parameters and a freely selected number of plot points \((x,F(x))\) of a nonlinear characteristic \(F(.)\) is proposed. It is assumed that the nonlinearity is bounded and vanishes at the origin. The identification routine is parametric and based on the idea of polynomial interpolation of the plot points \(\{(x,F(x)\}\), recursive least squares/prediction error method for estimation of system parameters under, respectively, bounded and ARMA-type noise, and a singular value decomposition for separation of the obtained aggregated identification outcomes. For the bounded noise case, the estimation errror upper bound is established, and for a noise-free case convergence of the resulting estimates is concluded assuming a particular, impulse-type, system input. Simulation illustrative examples support the theory.




Cited in
(54)








This page was built for publication: System identification based on Hammerstein model

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5460611)